Comparative Study of Machine Learning Algorithms and Correlation Between Input Parameters

Abstract

The availability of big data and computing power have triggered a big success in Artificial Intelligence (AI) field. Machine Learning (ML) becomes major highlights in AI due to the ability of self-improved as it is fed with more data. Therefore, Machine Learning is suitable to be applied in financial industry especially in detecting financial fraud which is one of the main challenges in financial system. In this paper, 15 different types of supervised machine learning algorithms are studied in order to find the highest accuracy that should be able to detect credit card fraudulent transactions. The best algorithm among these algorithms is then further used and studied to find the correlation between the input variables and the accuracy of the results produced. The results have shown that Multilayer Perceptron (MLP) produced the highest accuracy among the 15 other algorithms with 98% accuracy of detection. Besides that, the input parameters also play an important role in determining the accuracy of the results. Based on the result, when input parameter known as ‘V4’ decreased, the recorded accuracy has increased to 99.17%.